Calling Polkadot Core Developers

Their domination is projected to last for ~14 years.

Domination of what? The fellowship, as far as I understand, only has a single governance role, which is the ability to “whitelist” certain proposals hashes. As stated in that link:

That all said, oraclising well-informed opinion and allowing it to help optimise the decision-making process, even if it has no direct effect on the decision-making outcome seems like a reasonable goal to strive for. Crucially, and for the sake of all involved, it must not be possible in any way for the expert body to subvert the overall stakeholder decision.

So the fellowship will really have no special on-chain permissions over the decision-making outcome of proposals.

IMO the only argument to be made about the fellowship is whether you support a meritocracy or not, and it makes sense to me that the decisions about the direction and development of the Polkadot protocol should be in the hands of those who understand the technology best.

1 Like

Domination (in terms of voting power) in a fictional case where the Fellowship suddenly finds itself voting over a controversial question. I find it interesting because I think the Fellowship, or at least its governance design, could be used for decentralised decision making, as stated in the Manifesto. I explain more in the repo.

I think the Fellowship claims to be decentralised, so an interesting question is whether it is decentralised or is on a trajectory to becoming decentralised. Meritocracy is a larger and much more complicated question that I don’t want to have an opinion on here. I also don’t want to have an opinion on whether I support the Fellowship idea or if decentralisation is good or bad in this case. Just want to measure the Manifesto and the Fellowship on its own stated goals.

1 Like

Ultimately any sort of soft human consensus can only operate functionally under some degree of good-faith - while an analysis of a worst case of bad-faith operation is interesting I don’t believe it is realistic. Nevertheless, the power of the fellowship is strongly checked against such bad-faith operation.

I don’t think my analysis covers worst case or bad-faith. I was aiming for most likely case, assuming good faith operation, and visualizing the data & equations found in the seed list and the manifesto.

Is there something in the analysis I can change to make it seem more realistic? I thought I set the parameters and assumptions to very conservative values already.

Maybe the plotted scenario of Parity and Others voting 100% contrary to each other can be considered a worst case. I think it’s under-counting rather than over-counting Parity, and by extension Parity’s leadership’s, influence since there are rules in the manifesto that forbid lower ranks to vote too differently from what higher ranks do, if higher ranks agree. Parity will make up the higher ranks for many years.

Maybe the scenario of controversy can be considered a worst case. I agree with this, but a governance tool’s job is to continue producing decisions under worst case circumstances. Controversy is a common tough circumstance in crypto, so the Fellowship should be prepared for and benchmarked against this scenario. Any governance design works well during times when everyone agrees with their leaders.

On the topic of strongly checked we’re back to the good faith assumption quickly because the same people appear in all roles of Polkadot’s governance. DOT holders, council members, etc. This is not something these same people can easily change on their own though. Other external people simply have to take initiative and get involved on their own behalf, and it would be interesting to many if the Fellowship provided a framework or a green field for them to do so.

I’m not seeing bad faith in either the analysis or in the Fellowship. It looks like the Technical Committee is being replaced by a more involved but in practice (in terms of who makes decisions) identical governance body. They are the people who got Polkadot where it is today, and they’ll likely be as good faith in the Fellowship as they were in the TC.

It looks like the Fellowship might decentralise away from its initial top 3 members, and eventually away from Parity. So the claim of decentralisation isn’t entirely false or bad faith, it’ll just be slow. Maybe slower than intended?

I’m curious about what parameter or assumption in the simulation is unrealistic. The code should be fairly quick to search through. Please ask questions if you’re finding some part of it odd or hard to understand.

A proper worst case scenario would be what @rphmeier described earlier, where fellows disregard admission rules in order to keep outsiders out. Another version of that scenario would be if fellows disregard admission rules in order to get their own into higher positions quicker than rules allow.

This second version is what’s actually playing out, and it’s worse (in terms of decentralisation) than what I had modelled. See eg this PR where bkchr is awarded rank 6 despite obviously not fulfilling the criteria stated in the manifesto. There being no opposition to this in the public thread, and him referencing offline conversation underscores the level of centralisation and how close he is to his own approvers. They’re letting an insider in the back-door, which is what any game theoretic model would predict given the Fellowships design and structure.

During the initial bootstrapping phase some relaxations of these rules may be made in order to expediate decentralisation and ensure pluralism as early as possible in the member judgement process. Specifically, it may be reasonable to have shortened voting timelines and remove the need for some peripheral, demonstrative requirements if and only if it is obvious that the candidate in question possesses the relevant technical knowledge. In all cases, time limits on ranks should still generally be respected.

That’s taken from the manifesto. @bkchr is the top committer on both Substrate and Polkadot, I don’t think his technical knowledge is questionable. The role of the technical fellowship is to make technical decisions and therefore the people that are part of it should have the prerequisite knowledge. I don’t think it’s unexpected or controversial that the people who wrote most of the code so far are the ones that have the most knowledge and therefore have the highest rankings. We also shouldn’t “blindly” promote people just for the sake of “decentralization” since that detracts from the goal of being able to make good decisions.

IMO this is the most important aspect about the future of the fellowship which is not accounted by your model. The future health of the fellowship will be correlated with how fast we can grow the technical knowledge in the wider community, i.e. if we just sit around and wait for 14 years I can guarantee you that “Parity-adjacent” developers will still have the highest rankings in the fellowship. So in my view, spreading and growing the technical knowledge is the most important thing to be tackled, which will then lead to secondary effects such as: more people getting into the fellowship and getting promoted to higher ranks, thereby increasing its diversity.

I think the model might be useful to figure out if we can tune the transitioning terms or come up with “fast-tracks” for graduation, e.g. if someone writes a complete Polkadot node from scratch in a year, should they still have to wait another year to get to Dan III?

Lastly, I personally am OK with “fuzzy” rules, I don’t think you can come up with a set of rules that will capture all scenarios properly and I also don’t think we should throw away the capability of a group of humans coming together and making reasonable decisions. The power of the fellowship is limited to “fast-tracking” proposals that still need to be voted by the wider community, so the damage it could cause is not unbounded.

2 Likes

Thanks for the answer and for quoting the manifesto. That paragraph goes contrary to the decentralisation goals stated elsewhere in the manifesto, but I guess any organisation needs to strike a balance/prioritization between mutually exclusive goals. Here’s some example quotes anyways:

I’m not questioning bkchr’s knowledge, just for the record.

Absolutely agree. The model only considers the trajectory of the claimed decentralisation. I’m not qualified to have opinions about or modeling what the most important aspect is. I assume such discussions were held between experts before the Manifesto was published and acted upon? My first perspective on the Manifesto came in this thread. It was then alreadya finished text which put decentralisation front and center.

Even the part you quoted allowed rule relaxation only to “expediate decentralisation and ensure pluralism”, not to reward extra knowledgeable people.

The model assumes Polkadot is able to recruit a net surplus of 15 external core devs per year for 30 years straight, and that these people progress through the ranks equally fast as Parity devs. So there’s an assumption baked in, that you’re working extremely hard on spreading knowledge, which I already know you do.

How to define a good decision is a very tough problem within governance. The manifesto stays away from trying to define that from what I can find, but places big emphasis on decisions being decentralised. I think my gut feelings around this question are similar to yours, but again this is something that should have been discussed before the manifesto was published and acted upon.

This is very heavily criticised in the manifesto. The discussion that maybe belongs as an issue ticket in the manifesto repo.

Thanks! I really hope it can be useful to make some projections and make priorities visible and concrete.

2 Likes

Yes, I know it was just an example but I also wanted to provide some context to justify that decision.

The reason the process got expedited was the proven technical track record of that person, in that sense they are being rewarded for being knowledgeable (even though that wasn’t the main goal).

It’s a bit hard to come up with numbers for this (maybe someone can make a reasonable estimate) but if you account for all of the devs joining the ecosystem to work for multiple parachain teams and people working on alternative Polkadot implementations I’m sure that number was significantly higher than 15 in 2022 alone. Also there is a fair amount of churn at Parity as people leave the company to potentially focus on other Polkadot-related initiatives (or just leave the ecosystem altogether), I expect this to keep happening in the future as well.

I agree, trying to define “good” is an open philosophical question that we shouldn’t pursue. But this goes back to my earlier point that we shouldn’t try to codify everything. Also my understanding is that the manifesto is a “permanent work in progress”, it will be refined and adjusted over time, as we encounter new situations that lead to new learnings, or simply as norms and “traditions” evolve organically.

From the manifesto:

In the present work, our domain—and with it our problem surface—is intentionally constrained to the field of engineering core blockchain technology. Nonetheless, we are not yet at a point where we can define all judgements required by Fellowship voters in purely objective terms. Those members called upon to evaluate an individual will, to a greater or lesser extent need to make a judgement call. Pluralism, discussion and a clear framework for gathering a perspective is our antidote to the subjectivism which this otherwise creates.

The manifesto seems to mostly be in agreement with the fact that we are not able to create a framework that completely encapsulates the decision making process, there is room for subjectivity and we will defer to the best judgment of all individuals.

Being completely honest with you there are some questions you have for which none of us have definite answers, this is an on-going experiment in governance and only time will tell if we can reach the goals we set out to. In the immediate future the goal of the technical fellowship is to replace the council and technical committee. The council currently has 13 members and the technical committee has 3, in contrast the fellowship already has 37 members of which 20 are not currently employeed by Parity (and this isn’t a number we’re happy to settle on!), granted the highest ranking ones are mostly Parity employees due to the reasons I alluded to in my earlier post. The fellowship also will have significantly less power than the current council does. I would argue that this is already better, without having to quantify how much better, or how “decentralised” it is.

I hope you didn’t take my response as adversarial, it wasn’t meant that way.

No worries I’m assuming good intentions in all posts here.

I think I understand the Fellowship better now, having talked to many members. There are large playful, experimental and indeed centralized aspects of the whole idea. Different members have different ideas. The play and experimentation should be yours to enjoy and do what you like with anyways.

Mixing in words like decentralised and governance may attract scrutiny from many different external entities as the network grows, but that could be part of the fun if kept at a level you’re prepared for.

I hope the manifesto evolves gracefully and that you find a nice sweet-spot between fun and serious. Best of luck!

2 Likes

Just to be clear: one partial design goal of the fellowship is to be more decentralised than the current TC. This is not an idealistic goal, but rather a means to an end, most of which are already detailed in the manifesto and, generally revolve around safeguarding and growing the technical expertise required to keep the Polkadot network alive. Decentralisation helps mitigate the threat posed by a would be attacker or malign circumstances.

But “decentralisation” per se will not be enough to fulfil this mission. Taken to the extreme - handing one fellowship vote for each of the 8billion people on the planet - will not (I assert) result in a great outcome for this mission: almost all of those 8b people have no idea about Polkadot’s technicals. So the Fellowship recognises technical expertise in a sort of meritocracy in order to try to build and maintain this body of technical expertise but aims to decentralise as a matter of protecting against some risks that might otherwise be born from it.
I do not think that any of the Manifesto contradicts this viewpoint. While the part quoted does mention decentralisation, it is part of a wider assertion that openness, transparency and clarity of rules are crucial elements in any public governance system. As always these must be balanced with practicality and realistic aims. Idealism, absolutism and myopia rarely play well with social structures. In short, I don’t think it’s reasonable make a point of “decentralisation was mentioned and this is not wholly decentralised therefore there’s something wrong here”.

And, at the very least, it is substantially more open, transparent and inclusive than the current TC. It directly allows individuals outside of Parity and Web3 to have a say in the technical decision making and provides a path to ensure that, long-term, neither Parity nor the foundation are needed for the practical operation of the network.

1 Like

More specific terms would help discussion here. Let’s interpret political decentralisation three different ways. Since being more decentralised than the current TC is a partial design goal, I’ll evaluate it for each interpretation.

I1: “Many independent people are allowed to vote.”

This one is less useful than one might think. Eg it leaves the Fellowship vulnerable to straw-man arguments like the one you mentioned about 8 billion people.

I1-Evaluation: The TC had 3 voters (Parity, Parity, and W3F). The Fellowship has 37 members. 37 > 3.

I2: “Many independent people can reasonably expect to sometimes win referendums over disagreements.”

The word disagreement is important here. If there’s no disagreement, there should be no referendum, or else Governance is simply wasting energy, and people will not care to vote, which will undermine the governance mechanism.

I2-Evaluation: Let’s assume there were at least 3 independent people in the TC. All of them might sometimes win, sometimes loose a disagreement. There’s only one disagreement scenario: 2 winning side, 1 loosing side.

The top 6 members of the Fellowship have 115 votes. Everyone else (the 31 bottom members) combined have 104 votes. So superficially it might look like a minimum of 6 top members must agree to be sure to win a referendum. Since 6 > 2 this would arguably be more I2-decentralised than TC.

However, a scenario where one top 6 voter combined with a unison bottom 31 of voters, beats a unison top 5 of voters, is incredibly unlikely in the Fellowship’s system. Coordinating >20 voters is already as hard as it is unlikely, as shown by our political parties. In the Fellowship, the underdog-side is 32 actors, and the top 25 of them would have to vote in unison to win a referendum. Any leverage that top voters have on bottom voters would easily shrink the real practical number of top voters required below 6.

Such leverage exists in the Fellowship because members are not allowed to vote freely.

The highest rank among the everyone else group, and the lowest rank in the top 6 group, are both rank 5. The four rank 5 members must vote in line with the top 3 members (if those 3 agree) 80% of the time. The three rank 4 members must vote in agreement with superiors 90% of the time. The rest (27 members) must do so 100% of the time.

This non-free votes rule implicitly makes sure that the highest ranked members vote first, and in public. Otherwise, one can’t make sure to vote in line with them. Whichever rank 4 or 5 member who steps out of line first use their tiny fraction of free voting power, and make a big gamble that almost every voter after themselves will agree to oppose the top most qualified and authoritative members.

In practice this will almost never happen, and when it happens it will almost never succeed. We can predict this because British political parties are ruled similarly, with party discipline enforced by whips. The Fellowship will be ruled as one political party, with one whip, not many different ones. Parties, and the Fellowship, are designed to never split along the middle, and to demote problematic rare policy line breakers.

Almost all referendums will in practice be over when the top voter (rank 7) has discussed with the two rank 6 members. The three have enough leverage on the rest (34 members) that I expect them to never loose a single vote if they agree internally.

The discussion among the top 3 can be made very short in 70% of the disagreements, since the two rank 6 members must align with the one rank 7 member 70% of the time. You (Gavin) is the only member who can vote freely according to your rules written in the Manifesto.

Even if the TC had only 3 members, I’d say the Fellowship is potentially less I2-decentralised than the TC was, since the TC had no top voter with special powers. It had 3 actors who could vote freely, and 3 > 1.

The smallest group among the Fellowship required to never in practice loose a referendum might be 1, 2, 3, 4, 5, 6, or 7. It might compare favorably to the 2 observed in the TC. However, the coordination problem on the underdog-side is significantly worse in the Fellowship than it was in the TC (~25 > 2).

I3: “Governance can easily be untied from current leaders”

This one means that governance may continue uninterrupted even if some individuals suddenly disappear. An important consequence of it is that power will easily and peacefully be transferred every time.

I3-Evaluation: The Fellowship achieves this. However, the TC probably also achieved this, since companies voted, not individuals. Parity and W3F have their own mechanisms for transferring power and swapping out leaders.

The Fellowship ties voting power to individuals (sometimes for life) rather than to formal employment roles. You (Gavin) as the only Master member, will face no requirements, and can never be demoted, except by a general Polkadot referendum. Everyone else must defend their rank actively, and might get demoted much more easily.

There’s also something called “passive allowance” that will look suspicious to anyone who have experienced corruption. You wrote the rules, put yourself on top, are hard to remove, hard to overrule, and earn a passive income (from the Treasury?). As a Polkadot stakeholder I wish you had thought through the optics of this more closely before pubilshing.

This Master-4-life idea does not exist in normal companies, and hence not in the TC. It might be a step in the wrong direction in terms of I3-decentralisation.

Closing thoughts:
I think you walk into a self-contradiction with mentioning decentralisation as a goal for the Fellowship because so few qualify to a meaningful amount of power. Decentralisation among 1-7 people within the same company makes little sense for us outsiders.

I don’t know what made you use that word so much, but I hope it’s not idealism, absolutism or myopia. The Fellowship’s decentralisation, transparency and inclusiveness must be compared to Parity’s and W3F’s, since they were the voters in the TC. Parity and W3F already has some decentralised, transparent, and inclusive properties, and I’m not sure if the Fellowship as described will be more decentralised, transparent, or inclusive than Parity already was. The Fellowship might even be much worse, depending on how loosely you decide to interpret your own rules.

Governance design involves prioritizing between risks. I think it’s better to embrace that and speak clearly about it than to say there’s decentralisation here when there’s actually very little of it (except in the I1 sense).

A specification and classification of which threats, attacks, and malign circumstances we’re protecting against would be required to further our understanding of how the Fellowship is intended to work, and in which way it will be an improvement. Other than being a fun experiment.

1 Like

Noticed today that the polkadot.js.app displays Gavin’s rank as 6 instead of 7 as the seeding group Github repo suggested.

Screenshot from 2022-11-30 13-26-33

I can’t find the discussion that led to this change (searched Gavins Member request) and this forum. Regardless of the reasoning behind, seeding Gavin at rank 6 instead of 7 helps the Fellowship reach its goal of becoming more decentralised than the TC was. Some comments on how this changes the analysis in my previous post.

I1: No change.

I2: Gavin gets 7 fewer votes, and looses the 70% guaranteed influence on the two other rank 6 members. This improves everyone else’s chance of making their opinion count in a referendum.

The 3 > 1 special/free voters argument from my post above, is gone. It’s replaced by a more favorable 3 == 3. This makes the seeded version of the Fellowship equally or more I2-decentralised than the TC.

I3: The optics of “putting oneself on top” is improved as there are now 3 top members. Getting access to the discussion or reasoning which led to the 7->6 change would further improve these optics.

The change gives the Fellowship time and opportunity to abolish the Master-4-life idea if it wants, without having to potentially having to “fight” an existing and powerful Master in that process. (Assuming the Fellowship has the power to change the Manifesto. Getting access to discussions/referendums that’s behind Manifesto changes would further improve the optics of remaining I3-centralisation.)

As long as the Master’s benefits stay in place it’s possible that the Fellowship remains or becomes less l3-decentralised than the TC (Parity and W3F). That’s no problem if the probability is small. Moving Gavin from rank 7->6 increases the chances that the Fellowship is or will become more l3-decentralised than the TC.

2 Likes

Your analysis is really enlightening!
It’s great to see people getting interested enough to conduct such in-depth reviews on governance mechanism proposals.

There are a few reasons for the issues you pointed out, one being the mixing of status with operations, which let to master ranks being subject to much less constraints than lower ones (a status marker) while operation requirements and responsibilities should suggest the opposite (and this is effectively implemented for the lower ranks).

But the main part I cannot really figure out is about the rules governing the voting agreements between ranks. This is a part that I totally overlooked when I browsed through the manifesto so I am grateful that you pointed this out.
Usually, I would expect such coordination-enforcing rules to get baked-in in order to prevent attacks but I have not been able to find any justification that could make it relevant. The manifesto itself reads:

“A high degreee of disparity by a lower rank from a consensus of higher ranks implies either a material misinterpretation of this document which the lower rank should be pushed to correct, or some negligence in their voting routine which would imply a lack of commitment to the Fellowship’s norms.”

which is slightly weak as a justification.

As you mentioned, this set of rule totally impede the ability for lower ranks (in fact middle ones, but the exact span of this matters less than the principle) to vote independently.
Lower ranks then implicitly become nothing but a governance weight repository for the higher ranks, with the choice to either agree or get excluded. Abstaining being also punished, there is no real way to disagree with a master’s decision while remaining in the fellowship.

If the fellowship’s goal is to offer people with expertise in the protocol some way to technically impact the decision-making process, the peer-review mechanism through the ranking promotion/demotion already testifies for their expertise and legitimacy, and there should be no constraint in terms of voting like others.

Of course, there are other degrees of problems due to the fact promotions/demotions themselves are controlled by high ranking members, but that’s as second order issue compared to that one.

1 Like